tolk

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly instructs runtime fetching of documentation (e.g., curl -fsSL https://docs.ton.org/llms.txt and loading pages under https://docs.ton.org/tolk/), and those fetched pages are used as the authoritative source to guide the agent's prompts/behavior and are required for correct operation, so the external URLs directly control instructions at runtime.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill is specifically for building and operating Tolk smart contracts on The Open Network (TON) and explicitly covers wallets, Jettons (tokens), vesting, multisig, and composing/sending outgoing messages (createMessage(...) and .send(SEND_MODE_...)). It also uses a coins numeric type and addresses fee-sensitive paths and deployment/send behaviors. These are explicit crypto/blockchain financial operations (wallets, token handling, and on-chain sending), not generic tooling, so it grants direct financial execution capability.